Information processing device, data concealment support method, and data concealment support program

Abstract

Streamlining the process of data anonymization. [Solution] The information processing device comprises a data acquisition unit that acquires target data that may contain information to be kept confidential, and a target location estimation unit that estimates the parts of the target data that should be kept confidential using a machine learning-trained language model. This information processing device makes it possible to optimize the entire process related to data confidentiality.

Description

Technical Field

[0001] The present disclosure relates to an information processing apparatus, a confidentiality support method, and a confidentiality support program.

Background Art

[0002] Techniques for anonymizing a part of a document are known. As an example of a technique for anonymizing a part of a document, for example, a document processing apparatus described in Patent Document 1 below can be cited. This document processing apparatus divides the target document data into sentences, and determines whether each sentence is a target for anonymization based on a preset anonymization rule. Then, this document processing apparatus executes an anonymization process on the sentence determined to be a target for anonymization, and outputs document data including the sentence on which the anonymization process has been executed, that is, document data in which some sentences have been anonymized.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The document processing device described in Patent Document 1 determines whether each sentence is subject to anonymization based on pre-set anonymization rules. Furthermore, in the anonymization process, the document processing device anonymizes the target sentences based on pre-set anonymization rules. However, the information to be anonymized is generally diverse, and creating anonymization rules that can cover such diverse information is not easy. In particular, proper nouns, which are often the target of anonymization, consist of a vast number of existing proper nouns, as well as proper nouns used only in certain communities and newly created proper nouns. Therefore, it is extremely difficult to anonymize all proper nouns using the anonymization rules described in Patent Document 1.

[0005] Therefore, currently, it is necessary for people to manually check for any parts that need to be concealed, and the process of concealing data is not sufficiently efficient. Furthermore, the data subject to concealment is not limited to document data. For example, the same problems arise when concealing parts of image data (still or moving image data) or audio data.

[0006] This disclosure has been made in view of the above-mentioned issues, and one exemplary purpose is to provide a technology that enables the streamlining of data confidentiality work. [Means for solving the problem]

[0007] An information processing device relating to an illustrative aspect of this disclosure comprises a data acquisition means for acquiring target data that may contain matters to be kept confidential, and a target location estimation means for estimating the target locations in the target data that should be kept confidential, using a machine learning-trained language model.

[0008] Other information processing devices relating to illustrative aspects of this disclosure include: a target location identification means for identifying target locations in target data to be anonymized; and a mode estimation means for estimating a mode of anonymization to be applied to the target locations identified by the target location identification means, using a machine learning-trained language model.

[0009] In the example aspect of the anonymization support method relating to this disclosure, at least one processor performs a data acquisition process to acquire target data that may contain information to be anonymized, and a target location estimation process to estimate the target locations in the target data that should be anonymized, using a machine learning-trained language model.

[0010] An example of the concealment support program relating to this disclosure causes a computer to function as a data acquisition means for acquiring target data that may contain matters to be concealed, and a target location estimation means for estimating the target locations in the target data that should be concealed, using a machine learning-trained language model. [Effects of the Invention]

[0011] One exemplary effect of this disclosure is that it can provide a technology that enables the streamlining of data anonymization processes. [Brief explanation of the drawing]

[0012] [Figure 1] This is a block diagram showing the configuration of the information processing device related to this disclosure. [Figure 2] This is a flowchart showing the flow of the confidentiality support method related to this disclosure. [Figure 3] This is a block diagram showing the configuration of other information processing devices related to this disclosure. [Figure 4] This figure shows an example of a process for estimating the parts to be concealed. [Figure 5] This figure shows an example of a display screen for showing the parts to be concealed. [Figure 6]It is a diagram showing an example of a display screen for accepting selection or modification of conversion candidates. [Figure 7] It is a diagram showing an example of a process for estimating the mode of anonymization to be applied. [Figure 8] It is a diagram showing an example of a display screen for allowing selection of the mode of conversion to be applied. [Figure 9] It is a diagram showing another example of a display screen for allowing selection of the mode of conversion to be applied. [Figure 10] It is a flowchart showing the flow of a process for identifying parts to be anonymized. [Figure 11] It is a flowchart showing the flow of a process for generating anonymized data. [Figure 12] It is a block diagram showing the configuration of an information processing apparatus according to a reference example. [Figure 13] It is a block diagram showing the configuration of a computer that functions as an information processing apparatus according to the present disclosure.

Embodiments for Carrying Out the Invention

[0013] Hereinafter, embodiments of the present invention will be exemplified. However, the present invention is not limited to the following exemplary embodiments, and various modifications are possible within the scope shown in the claims. For example, embodiments obtained by appropriately combining the technologies (part or all of the objects or methods) employed in the following exemplary embodiments may also be included in the scope of the present invention. Also, embodiments obtained by appropriately omitting part of the technologies employed in the following exemplary embodiments may also be included in the scope of the present invention. Further, the effects mentioned in the following exemplary embodiments are merely examples of the effects expected in those exemplary embodiments and do not define the scope of the present invention. That is, embodiments that do not exhibit the effects mentioned in the following exemplary embodiments may also be included in the scope of the present invention.

[0014] 〔First Exemplary Embodiment〕 A first exemplary embodiment, which is an example of an embodiment of the present invention, will be described in detail with reference to the drawings. This exemplary embodiment is a basic form for each of the exemplary embodiments described later. Note that the scope of application of each technology adopted in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technology adopted in this exemplary embodiment can also be adopted in other exemplary embodiments included in this disclosure as long as there is no particular technical problem. In addition, each technology shown in the drawings referred to for explaining this exemplary embodiment can also be adopted in other exemplary embodiments included in this disclosure as long as there is no particular technical problem.

[0015] (Configuration of Information Processing Apparatus 1) The configuration of the information processing apparatus 1 according to this exemplary embodiment will be described with reference to FIG. 1. FIG. 1 is a block diagram showing the configuration of the information processing apparatus 1. As shown in FIG. 1, the information processing apparatus 1 includes a data acquisition unit 101 and a target location estimation unit 102.

[0016] The data acquisition unit 101 acquires target data that may include items to be anonymized. Here, anonymization means making the content unknowable. For example, processes such as deleting parts that are not desired to be known and performing masking processing so that the parts cannot be viewed are also included in the scope of anonymization. Anonymization can also be referred to as, for example, pseudonymization. Note that hereinafter, performing masking processing will be referred to as masking. In addition, the process of replacing parts that are not desired to be known with content that can be known without problem will also be described as an aspect of masking.

[0017] The target data described above can be any data that may contain information that should be kept confidential, and any electronic data can be used as the target data. For example, the target data may be document data (which can also be called text data), image data (still image data or moving image data), or audio data. Furthermore, the content of the target data is not particularly limited. For example, the target data may be an internal company document that is scheduled to be made public, an audio recording of a meeting, or document data such as a news draft that will be publicly released. Alternatively, for example, the target data may be a medical document such as an electronic medical record. Thus, the information processing device 1 can also be applied to the healthcare field. Furthermore, the method of acquiring the target data is also arbitrary; for example, the data acquisition unit 101 may acquire the target data input to the information processing device 1, or it may acquire the target data stored in the information processing device 1 or another device.

[0018] The target location estimation unit 102 estimates the target locations in the target data that should be anonymized, using a pre-trained language model. Hereinafter, the language model used by the target location estimation unit 102 will be referred to as language model M.

[0019] Language model M is a model that has been trained on natural language. More specifically, training on natural language means learning the arrangement of its constituent elements (such as words) in natural language sentences and the arrangement of sentences in texts. A language model M that has been trained on natural language can output information useful for estimating appropriate target locations in accordance with the context of the target data. Examples of language models that have been trained on natural language include BERT (Bidirectional Encoder Representations from Transformers), RoBERTa (Robustly optimized BERT approach), and ELECTRA (Efficiently Learning an Encoder that Classifies Token Replacements Accurately). Language model M may be a general-purpose model that can be used for various purposes, a general-purpose model that has been fine-tuned for estimating target locations for anonymization, or a dedicated model trained specifically for estimating target locations for anonymization.

[0020] Furthermore, estimating the target location using the language model M means directly or indirectly utilizing the language model M in estimating the target location. There are no particular limitations on how the language model M is utilized in estimating the target location. For example, if the target data is text data written in natural language, the target location estimation unit 102 may input the target data directly into the language model M and have it infer the target location in the input target data.

[0021] Furthermore, if the target data is in a non-text format, the target location estimation unit 102 may convert the data into text format and input it to the language model M. For example, various well-known methods such as speech recognition technologies and OCR (Optical Character Recognition) can be used to convert audio data into text format, and to convert image data into text format. It is also possible to convert image data into text format using a language model trained to take image data as input and output the content of that image data. Note that if the target data is image data, it is not necessarily required to convert the entire image into text. For example, the target location estimation unit 102 may estimate the area to be concealed from the text obtained by performing OCR on the character portion of the image data.

[0022] Furthermore, some language models are configured and trained to accept non-textual data such as image data. When the target location estimation unit 102 uses a language model M that can accept non-textual data, it can directly input the non-textual target data into the language model M.

[0023] The target location estimation unit 102 may either use the output of the language model M as the estimation result, or it may estimate the target location based on the output of the language model M. In the former case, the target location estimation unit 102 can simply have the language model M infer the target location. In the latter case, the target location estimation unit 102 may, for example, have the language model M output candidate target locations and a confidence level indicating the likelihood that each candidate is a target location, and estimate the candidates whose confidence level is above a predetermined threshold as target locations.

[0024] Furthermore, the language model M may be stored in the information processing device 1, or it may be stored in a server or the like outside the information processing device 1. In the latter case, the target location estimation unit 102 can use the language model M via the server or the like that stores the language model M.

[0025] As described above, the information processing apparatus 1 according to this exemplary embodiment employs a configuration comprising: a data acquisition unit 101 that acquires target data that may contain matters to be made confidential; and a target location estimation unit 102 that estimates the target locations in the target data that should be made confidential using a machine learning-trained language model M.

[0026] According to the above configuration, since the language model M is used to estimate the parts of the target data that need to be anonymized, it is possible to estimate the parts that need to be anonymized according to the content of the target data without having to create anonymization rules like those explained in [Background Technology] in advance. This has the effect of eliminating or reducing the burden of manually checking the parts that need to be anonymized, thereby making data anonymization work more efficient. Furthermore, by using the information processing device 1, it is also possible to optimize the entire data anonymization process.

[0027] Furthermore, how the target locations estimated by the target location estimation unit 102 are used in data anonymization is at the user's discretion. For example, the information processing device 1 may present the estimated target locations to the user. In this case, the user can decide whether or not to anonymize the presented target locations. Alternatively, for example, the information processing device 1 may automatically anonymize the estimated target locations and generate anonymized data. In this case, it becomes possible to automatically generate anonymized data from the target data without creating the burden of manually verifying the target locations.

[0028] (Anonymization Support Program) The functions of the information processing device 1 described above can also be implemented by a program. The data concealment support program according to this exemplary embodiment causes the computer to function as a data acquisition means for acquiring target data that may contain information to be concealed, and a target location estimation means for estimating the target locations in the target data that should be concealed, using a machine learning-trained language model M. This data concealment support program has the effect of making data concealment work more efficient.

[0029] (Flowchart of the confidentiality support method) The flow of the data concealment support method according to this exemplary embodiment will be explained with reference to Figure 2. Figure 2 is a flowchart showing the flow of the data concealment support method. Note that the entity executing each step in this data concealment support method may be a processor provided in the information processing device 1, a processor provided in another device, or the entity executing each step may be a processor provided in a different device.

[0030] In S1 (data acquisition process), at least one processor acquires target data that may contain information that should be kept confidential.

[0031] In S2 (target location estimation process), at least one processor uses a machine learning-based language model M to estimate the target locations in the target data acquired in S1 that should be anonymized.

[0032] As described above, the data anonymization support method according to this exemplary embodiment employs a configuration in which at least one processor performs a data acquisition process to acquire target data that may contain information to be anonymized, and a target location estimation process to estimate the parts of the target data that should be anonymized using a machine learning-trained language model M. This data anonymization support method has the effect of making data anonymization work more efficient.

[0033] [Second exemplary embodiment] A second exemplary embodiment, which is an example of an embodiment of the present invention, will be described in detail with reference to the drawings. Components having the same function as those described in the above-described exemplary embodiment are denoted by the same reference numerals, and their descriptions are omitted as appropriate. The scope of application of each technology adopted in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technology adopted in this exemplary embodiment can also be adopted in other exemplary embodiments included in this disclosure, to the extent that no particular technical problems arise. Furthermore, each technology shown in the drawings referenced to describe this exemplary embodiment can also be adopted in other exemplary embodiments included in this disclosure, to the extent that no particular technical problems arise.

[0034] (Configuration of Information Processing Device 1A) The configuration of the information processing device 1A according to this exemplary embodiment will be described with reference to Figure 3. Figure 3 is a block diagram showing the configuration of the information processing device 1A. The information processing device 1A is a device equipped with a function to support data confidentiality. The information processing device 1A may be a local device used by individual users, or it may be a server that provides a service to support data confidentiality to multiple users.

[0035] As shown in the figure, the information processing device 1A includes a control unit 10A that controls all parts of the information processing device 1A, and a storage unit 11A that stores various data used by the information processing device 1A. The information processing device 1A also includes a communication unit 12A for the information processing device 1A to communicate with other devices, an input unit 13A that receives input to the information processing device 1A, and an output unit 14A for the information processing device 1A to output data. The control unit 10A includes a data acquisition unit 101A, a target location estimation unit 102A, a reception unit 103A, a target location identification unit 104A, a pattern estimation unit 105A, a concealment unit 106A, and a presentation control unit 107A.

[0036] The data acquisition unit 101A acquires target data that may contain information to be kept confidential, similar to the data acquisition unit 101 in the exemplary embodiment 1. In this exemplary embodiment, an example is described in which the target data is document data in text format. For example, the data acquisition unit 101A may acquire target data input by a user of the information processing device 1A, i.e., the person who will make the target data confidential, via the input unit 13A. Alternatively, the data acquisition unit 101A may acquire target data from another device (e.g., a terminal device owned by the user) via the communication unit 12A. Furthermore, the target data may be stored in the storage unit 11A or an external storage device of the information processing device 1A, in which case the data acquisition unit 101A simply needs to access the storage unit 11A or the storage device to acquire the target data.

[0037] The target location estimation unit 102A, similar to the target location estimation unit 102 in the exemplary embodiment 1, estimates the target locations in the target data that should be anonymized, using a machine learning-trained language model M. The method of estimating the target locations by the target location estimation unit 102A will be described in detail later.

[0038] The reception unit 103A receives various instructions from the user. As will be described in detail later, the reception unit 103A receives, for example, the input of information to be kept confidential, the specification of the parts to be kept confidential, etc. The method of receiving instructions is arbitrary. For example, the reception unit 103A may receive instructions via the input unit 13A, or it may receive instructions from other devices (for example, terminal devices used by the user) via the communication unit 12A.

[0039] The target location identification unit 104A identifies the target locations in the target data that are to be concealed. Specifically, the target location identification unit 104A identifies the locations that have been estimated by the target location estimation unit 102A and whose designation has been accepted by the reception unit 103A as the locations to be concealed. In other words, the target location estimation unit 102A estimates candidates for locations to be concealed, and the locations specified by the user from among the estimated candidates are identified as locations to be concealed by the target location identification unit 104A.

[0040] The mode estimation unit 105A estimates the mode of concealment to be applied to the target location identified by the target location identification unit 104A, using a machine learning-trained language model M. The mode of concealment only needs to make the content of the target location unrecognizable or difficult to recognize. For example, deleting the target location from the target data and masking the target location are modes of concealment. The mode estimation unit 105A may also estimate what kind of masking to apply as a mode of concealment. In this exemplary embodiment, an example of how the mode estimation unit 105A will transform the target location, in other words, an example of estimating candidate transformations for the target location, will be described.

[0041] The manner estimation unit 105A may use a different language model than the language model M used by the target location estimation unit 102A to estimate the manner of concealment to be applied to the target location. In that case, the target location estimation unit 102A may use a language model that has been finely tuned for estimating the target location, and the manner estimation unit 105A may use a language model that has been finely tuned for estimating the manner of concealment.

[0042] Furthermore, as described in Exemplary Embodiment 1, the target data may be non-text data such as image data or audio data. If the target data is non-text data, the manner estimation unit 105A should estimate the manner of concealment according to its format. For example, if the target data is image data, it can be concealed by masking, so the manner estimation unit 105A should estimate the manner of masking to be applied to the target area of ​​the image data. For example, the manner estimation unit 105A may estimate whether to mask the target area with black paint, mask it with mosaic processing, or mask it by superimposing another image. Also, for example, if the target data is audio data, it can be concealed by superimposing another audio onto a part of it, so the manner estimation unit 105A should estimate the audio to be superimposed onto the target area of ​​the audio data.

[0043] Furthermore, as will be explained in more detail later, the manner estimation unit 105A can estimate multiple conversion candidates for a single target location. For example, if the location to be concealed is the place name "XYZ City" in the image data, the manner estimation unit 105A can estimate multiple conversion candidates such as "a certain city," "a certain major city," and "a certain city in the Tokyo metropolitan area." In this way, the information processing device 1A makes it possible to conceal non-text format target data, such as image data, in a manner that conforms to the user's intentions.

[0044] The concealment unit 106A conceals the target portion of the target data to generate concealed data. As described above, the target portion of the target data is the portion that the user has designated as the target for concealment from among the target portion estimated by the target portion estimation unit 102A. Furthermore, although the details will be explained later, the concealment method is determined based on the estimation result of the method estimation unit 105A.

[0045] Furthermore, if the target data is image data, a portion of the image data (for example, a portion showing a person's face) may be the area to be anonymized. In that case, the anonymization unit 106A may use an object detection model trained to detect areas in the image data in which a predetermined object is found to detect the location and extent of the target area in the target data, and generate anonymized data by masking that location and extent in the target data.

[0046] The presentation control unit 107A presents various information necessary for concealment support to the user of the information processing device 1A. For example, the presentation control unit 107A may present the target locations estimated by the target location estimation unit 102A. In that case, the reception unit 103A accepts the designation of the target locations to be concealed from among the target locations presented by the presentation control unit 107A. This allows for the smooth concealment of locations that align with the user's intentions.

[0047] The presentation control unit 107A has the discretion to decide what device to have the information presented and in what manner. For example, the presentation control unit 107A can present various types of information in any manner, such as display, printing, sound, or a combination thereof. For example, if the output unit 14A is a display device, the presentation control unit 107A may have the output unit 14A display and output various types of information.

[0048] As described above, the information processing device 1A, like the information processing device 1, includes a data acquisition unit 101A that acquires target data that may contain information to be kept confidential, and a target location estimation unit 102A that estimates the parts of the target data that should be kept confidential using a machine learning-trained language model M. Therefore, the information processing device 1A has the effect of making data confidentiality work more efficient.

[0049] Furthermore, as described above, the information processing device 1A includes a mode estimation unit 105A that estimates the mode of concealment to be applied to the target location using a language model M. This provides the added benefit of automating or semi-automating the determination of the mode of concealment, in addition to the effects of the information processing device 1, thereby further streamlining data concealment operations.

[0050] (Regarding the estimation of the parts that have been kept secret) The process by which the target location estimation unit 102A estimates the locations to be concealed will be explained with reference to Figure 4. Figure 4 is a diagram showing an example of the process for estimating the locations to be concealed. In the example in Figure 4, user U inputs text 41 indicating the matters to be concealed and target data 42, which is document data that may contain the matters to be concealed, to the information processing device 1A. The input target data 42 is acquired by the data acquisition unit 101A, and the text 41 is received by the reception unit 103A.

[0051] Text 41 contains information that user U wishes to conceal, expressed in natural language. Specifically, text 41 includes the sentences, "I want to conceal the company name," and "I also want to conceal 'Kawasaki City' and related information." In this way, the information processing device 1A (more precisely, the reception unit 103A) can accept input of multiple items to be concealed. The reception unit 103A may also accept input of information that abstractly indicates the items to be concealed, such as "I want to conceal the company name," or input of information that directly indicates the items to be concealed, such as "I also want to conceal 'Kawasaki City' and related information." For example, the reception unit 103A may accept input of any keyword to be concealed. In addition, the reception unit 103A may accept, for example, a specific company name, an individual's name, address, telephone number, email address, and confidential technical information as items to be concealed.

[0052] Furthermore, the reception unit 103A may accept input of information that indirectly indicates matters to be kept confidential. For example, the reception unit 103A may accept input of information indicating the intended use of the target data. This is because the matters to be kept confidential may vary depending on the intended use of the target data. To give a specific example, if the target data is a news draft, the reception unit 103A may accept input of information indicating that the target data will be used for reporting. This makes it possible to infer that the parts to be kept confidential are those that should be kept confidential when reporting (for example, descriptions that could identify an individual's address). Also, for example, if the target data is an internal company document, the reception unit 103A may accept input of information indicating that the target data will be made public. This makes it possible to infer that the parts to be kept confidential are those that contain confidential information that should not be leaked outside the company, or those that would cause compliance problems if made public.

[0053] Next, the target location estimation unit 102A estimates the target location in the target data 42 based on the input text 41. Specifically, in the example in Figure 4, the target location estimation unit 102A generates a prompt 43 instructing the system to extract the target location from the target data 42. The target location estimation unit 102A then inputs the generated prompt 43 into the language model M, which outputs a response 44 indicating the target location in the target data 42.

[0054] Prompt 43 instructs the user to extract all the parts of the target data 42 that correspond to the information to be kept confidential and to provide the extracted parts as the answer. Prompt 43 also includes the text of the target data 42 and the text of text 41. Since everything in Prompt 43 except the text of the target data 42 and text 41 is standardized, these standardized parts can be templated in advance. In that case, the target area estimation unit 102A can generate Prompt 43 by inputting the target data 42 and text 41 into the template.

[0055] Response 44 indicates "XXX Food Service" and "Kawasaki City" as the target locations. "XXX Food Service" was extracted based on the statement in text 41, "I want to conceal the company name." On the other hand, "Kawasaki City" was extracted based on the statement in text 41, "I also want to conceal 'Kawasaki City' and information related to it."

[0056] The target location estimation unit 102A only needs to generate a prompt that can output the information necessary to estimate the target location, and the prompt is not limited to those exemplified in Figure 4. For example, if no information to be concealed is entered, the target location estimation unit 102A may generate a prompt that does not include the information to be concealed. In this case, the target location estimation unit 102A may generate a prompt with text such as, "Infer whether the following target data contains information that should be concealed, and if so, extract that information and provide your response."

[0057] Furthermore, the target location estimation unit 102A may generate prompts containing various information to improve the accuracy of estimating target locations in line with the user's intent. For example, the target location estimation unit 102A may generate a prompt that includes anonymized sample data showing data that has been partially anonymized in the past and the anonymized parts in that data, and instructs the system to extract the parts of the target data 42 that should be anonymized by referring to the anonymized sample data. This makes it possible to extract target locations using the same criteria as the anonymized sample data.

[0058] Furthermore, when a keyword to be concealed is entered, the target location estimation unit 102A may generate a prompt instructing it to extract locations related to the keyword from the target data 42 and output a relevance score, which is an index value indicating the degree of relevance between the keyword and the extracted locations. In this case, the presentation control unit 107A can present multiple target locations related to a single keyword in order of their relevance to that keyword.

[0059] Furthermore, the target location estimation unit 102A may generate a prompt instructing it to extract the locations to be concealed from the target data 42 and to output the reason for extracting those locations. In this case, the presentation control unit 107A can present the target locations along with the reason for their extraction.

[0060] As described above, the information processing device 1A includes a reception unit 103A that receives input of matters to be concealed, and the target location estimation unit 102A may generate a prompt that instructs the extraction of the locations corresponding to the above matters from the target data. The target location estimation unit 102A may then estimate the locations to be concealed based on the output obtained by inputting the generated prompt into the language model M. This provides the effect that, in addition to the effects of the information processing device 1, it becomes possible for the user to estimate the locations corresponding to the matters they wish to conceal as target locations.

[0061] (Example of a display screen that allows the user to select the areas to be concealed) As described above, the presentation control unit 107A may present the target locations estimated by the target location estimation unit 102A to the user. Figure 5 shows an example of a display screen for presenting the locations to be concealed. In the example screen Img1 shown in Figure 5, the target locations estimated by the target location estimation unit 102A are displayed in a list, and each target location is displayed with a corresponding checkbox 51. In addition, the example screen Img1 displays text prompting the user to check the target locations they wish to conceal.

[0062] By displaying a screen like the example Img1, the user can quickly identify the parts of the target data that are considered to require concealment. Furthermore, the user can specify the target areas with a simple operation, such as manipulating the cursor 52 to check the checkbox 51. The target areas specified by the user are then identified as areas to be concealed by the target area identification unit 104A.

[0063] Furthermore, in example screen Img1, the target areas are grouped according to the information the user wishes to conceal. In other words, in example screen Img1, the target areas grouped as "Company Name" are those extracted based on the statement in text 41 shown in Figure 4, "I want to conceal the company name." Similarly, in example screen Img1, the target areas grouped as "Related to 'Kawasaki City'" are those extracted based on the statement in text 41 shown in Figure 4, "I also want to conceal 'Kawasaki City' and related information."

[0064] Furthermore, in the example screen Img1, the cursor 52 is positioned over the word "Tamagawa," which is one of the target locations, and as a result, object 53 is displayed indicating the reason why the word "Tamagawa" was extracted as a target location. In this way, the presentation control unit 107A may present the reason why a particular target location was extracted in response to an operation to select a presented target location. This allows the user to decide whether or not to specify a target location by referring to the presented reason for extraction. As mentioned above, the reason for extraction of a target location can be output to the language model M.

[0065] As described above, the information processing device 1A includes a presentation control unit 107A that presents the target locations estimated by the target location estimation unit 102A, and a reception unit 103A that receives the designation of target locations to be concealed from among the target locations presented by the presentation control unit 107A. This provides an additional benefit to the information processing device 1, as well as the ability to designate locations that align with the user's intentions from among the target locations estimated by the target location estimation unit 102A as targets for concealment.

[0066] (Example of a display screen that accepts selection or modification of conversion candidates) After the areas to be concealed are identified as described above, the manner estimation unit 105A estimates the manner of concealment to be applied to the identified areas. Then, as described above, the manner estimation unit 105A may estimate conversion candidates for the areas to be applied to. The presentation control unit 107A may present the conversion candidates estimated by the manner estimation unit 105A to the user. Figure 6 shows an example of a display screen that accepts the selection or modification of conversion candidates.

[0067] In the example screen Img2 shown in Figure 6, the target locations identified by the target location identification unit 104A are displayed in a list, and each target location is associated with a conversion candidate. In addition, each target location is grouped according to the information that the user wishes to conceal, similar to the example screen Img1 in Figure 5. For example, in Img2, the target location "XXX Food Service" is displayed in the "Company Name" group. A text box 61 displaying the conversion candidate "Company_1" is then associated with this target location. Note that the conversion candidates for each target location do not necessarily need to be estimated by the pattern estimation unit 105A; they may be determined by a rule-based system or the like.

[0068] Furthermore, the example screen Img2 displays a message prompting the user to confirm whether or not to convert the target area to the conversion candidates displayed in text box 61, and to press the confirmation button 62 if conversion is possible. The confirmation button 62 is a software key. When the reception unit 103A receives an operation to select the confirmation button 62, the concealment unit 106A decides to convert each target area to the conversion candidate displayed in each text box 61. The process of converting the target area to the conversion candidate can also be described as the process of replacing with the conversion candidate, or the process of masking the target area with the conversion candidate, etc.

[0069] Furthermore, the example screen Img2 also displays a message prompting the user to rewrite the text if they wish to change the converted description, i.e., the conversion candidates. In other words, in the example screen Img2, the text displayed in text box 61 is editable by the user. After the text displayed in text box 61 is modified, if the reception unit 103A receives an operation to select the confirmation button 62, the concealment unit 106A decides to convert each target area to the conversion candidate displayed in each text box 61, i.e., the modified conversion candidate.

[0070] As described above, the presentation control unit 107A may present conversion candidates for concealing the target area, and the reception unit 103A may accept the selection or modification of the presented conversion candidates. The concealment unit 106A may convert the target area in the target data into the selected conversion candidate to generate concealed data, or convert the target area in the target data into the modified conversion candidate to generate concealed data. In addition to the effects performed by the information processing device 1, the user can confirm how the target area will be converted into a conversion candidate, modify the presented conversion candidate as needed, and generate concealed data that conforms to their intentions.

[0071] (Regarding the presumption of the appropriate mode of secrecy) The outline of the process by which the mode estimation unit 105A estimates the mode of concealment to be applied will be explained with reference to Figure 7. Figure 7 is a diagram showing an example of the process for estimating the mode of concealment to be applied. In the example in Figure 7, a prompt 71 generated by the information processing device 1A (more precisely, the mode estimation unit 105A) is input to the language model M, and a response 72 to the prompt 71 is output from the language model M. Then, the concealed data 73 generated based on the response 72 is presented to the user U.

[0072] Prompt 71 instructs the system to infer the appropriate mode of anonymization to apply, more specifically, how the target portion of the target data should be transformed into anonymized descriptions, and then respond with transformation candidates. Prompt 71 also includes a statement instructing the system to respond with multiple transformation candidates with different degrees of abstraction for a single target portion. Including such a statement makes it easier to generate transformation candidates with varying degrees of abstraction, thereby facilitating the generation of anonymized data that aligns with the user U's intentions. The mode estimation unit 105A may also generate a prompt that specifies the direction or perspective of abstraction, in other words, how to abstract the data. For example, the mode estimation unit 105A may generate a prompt that instructs the system to generate multiple transformation candidates by abstracting each target portion in a different direction.

[0073] Furthermore, the prompt 71 includes each target location as well as the target data. The mode estimation unit 105A can generate such a prompt by inputting the target locations identified by the target location identification unit 104A and the target data acquired by the data acquisition unit 101A into a predetermined template. Alternatively, the target locations estimated by the target location estimation unit 102A may be input instead of the target locations identified by the target location identification unit 104A. In addition, it is not mandatory to include the target data in the prompt, but if the target data is included, the mode of concealment that should be applied to the target location can be estimated by considering the context of the target data. From the perspective of considering the context, it is not necessarily required to include the entire text of the target data in the prompt; at least the sentence containing the target location should be included in the prompt.

[0074] The mode estimation unit 105A only needs to generate a prompt that can output information necessary to estimate the mode of concealment to be applied, and the prompt is not limited to those exemplified in Figure 7. For example, the mode estimation unit 105A may generate a prompt that includes information indicating the matters to be concealed, which has been input by user U, and instructs the unit to estimate the mode of concealment to be applied by referring to that information. This makes it possible to infer an appropriate mode of concealment according to what matters user U wants to conceal.

[0075] Furthermore, for example, the mode estimation unit 105A may include conversion sample data showing the locations to be anonymized in data that have been anonymized in the past and the converted descriptions of those locations, and generate a prompt instructing the unit to estimate the mode of anonymization to be applied by referring to the conversion sample data. This makes it possible to infer the mode of anonymization to be applied using the same criteria as the conversion sample data.

[0076] Furthermore, the mode estimation unit 105A may estimate the mode of concealment to be applied and generate a prompt instructing it to output the reason for recommending that mode. In this case, the presentation control unit 107A can present the mode of concealment to be applied along with the reason for its recommendation.

[0077] In response 72 shown in Figure 7, multiple conversion candidates with different degrees of abstraction are presented for each target section. Specifically, for the target section "XXX Food Service," three conversion candidates are shown: "Company_1," "Emerging Company_1," and "Company in the Food Service Industry_1." All of these are abstractions of the specific company name "XXX Food Service," but they differ in their degree of abstraction. In other words, "Company_1" is the most abstract, and it is impossible to tell what kind of company it is from this description, while "Emerging Company_1" is specific enough to indicate that it is a newly established company, and "Company in the Food Service Industry_1" is specific enough to indicate the industry.

[0078] The anonymized data 73 shown in Figure 7 is data generated by converting the parts to be anonymized in the target data 42 shown in Figure 4 into the conversion candidates shown in Answer 72. More specifically, the conversion candidate used to generate the anonymized data 73 is the most abstract conversion candidate among the conversion candidates shown in Answer 72. The conversion candidate used to generate the anonymized data may be selected by the user U, or it may be automatically selected by the anonymization unit 106A.

[0079] (Example 1 of a display screen for selecting the type of conversion to apply) Figure 8 shows an example of a display screen for selecting the mode of transformation to apply. In the example screen Img3 shown in Figure 8, the target data is displayed. The parts of this target data that are to be concealed are highlighted. In addition, the example screen Img3 displays a preview image 81 of the concealed data generated when a transformation candidate with a high degree of abstraction is applied from among the multiple transformation candidates generated by the mode estimation unit 105A to the language model M, and a selection button 82 for selecting this transformation candidate is also displayed. Furthermore, the example screen Img3 displays a preview image 83 of the concealed data generated when a transformation candidate with a low degree of abstraction is applied from among the multiple transformation candidates generated by the mode estimation unit 105A to the language model M, and a selection button 84 for selecting this transformation candidate. In the preview images 81 and 83 of the concealed data, the transformed parts are highlighted. Note that both the selection buttons 82 and 84 are software keys.

[0080] Thus, the presentation control unit 107A may present the data after converting the parts of the target data to be concealed into conversion candidates generated by the language model M. In other words, the presentation control unit 107A may preview the concealed data that will be generated when the conversion candidates are applied. This allows the user to recognize what the concealed data will look like when the conversion candidates are applied, and then decide whether or not to apply the conversion candidates.

[0081] When a user determines that the concealment method shown in preview image 81 is preferable, they can select the selection button 82 displayed in conjunction with preview image 81. This generates the concealed data shown in preview image 81. Similarly, when a user determines that the concealment method shown in preview image 83 is preferable, they can select the selection button 84 displayed in conjunction with preview image 83. This generates the concealed data shown in preview image 83.

[0082] Furthermore, similar to the example in Figure 6, the reception unit 103A may accept user modifications to the concealment configuration presented by the presentation control unit 107A. This makes it possible to modify a portion of the previewed concealment configuration to align with the user's intentions.

[0083] (Example 2 of a display screen for selecting the type of conversion to apply) Figure 9 shows another example of a display screen for selecting the mode of transformation to apply. In the example screen Img4 shown in Figure 9, the target data is displayed in the same way as in the example screen Img3 in Figure 8. In the target data shown in the example screen Img4, the parts to be concealed are highlighted in bold, and the part selected by the cursor 91 (specifically the text "XXX Food Service") is highlighted with a mark. In this way, the presentation control unit 107A may present the parts to be concealed in the target data in a way that makes them distinguishable from other parts, and furthermore, it may present the parts to be concealed that the user has specified in a way that makes them distinguishable from other parts.

[0084] In addition, in the example screen Img4, the conversion candidates for the target area selected by the cursor 91 are displayed in a list in the display area 92. The user can select the conversion candidate to apply to the concealment of the target area selected by the cursor 91 from the list of conversion candidates and then select the confirmation button 93 to confirm that the target area has been converted to the selected conversion candidate. For example, if the user selects "Food Service Industry Companies_1" with the cursor 91 and then selects the confirmation button 93, it is confirmed that the description "XXX Food Service" in the target data will be converted to "Food Service Industry Companies_1".

[0085] Thus, the presentation control unit 107A may, in response to an operation to specify a target location, present a mode of concealment to be applied to that target location, which has been estimated by the mode estimation unit 105A. For example, as in the example in Figure 7, if the language model M generates multiple conversion candidates, each abstracting the target location at a different level of abstraction, the presentation control unit 107A may present the generated multiple conversion candidates as shown in the example screen Img4. The reception unit 103A may then accept a specification of the conversion candidate to be used for converting the target location from among the multiple conversion candidates presented by the presentation control unit 107A. The specified conversion candidate is used by the concealment unit 106A to generate concealed data.

[0086] Furthermore, the display control unit 107A may, depending on which of the multiple displayed conversion candidates is selected, display a preview image in which the target portion of the displayed target data has been converted to the selected conversion candidate.

[0087] Furthermore, the manner in which multiple conversion candidates are presented is arbitrary and is not limited to the example in Figure 9. For example, the presentation control unit 107A may display a list of each conversion candidate for each target location without requiring the user to select a target location with the cursor 91. Alternatively, as explained with reference to Figure 8, the presentation control unit 107A may present multiple conversion candidates by previewing the anonymized data that will be generated when each of the multiple conversion candidates is applied.

[0088] Furthermore, since the mode of concealment is not limited to conversion by conversion candidates, the presentation control unit 107A may present candidates for the mode of concealment to be applied (for example, deletion or masking of the target area) instead of presenting conversion candidates. Also, when the presentation control unit 107A presents candidates for the mode of concealment to be applied, it may also display a preview of the concealed data that will be generated when each of the multiple candidates is applied, as shown in the example in Figure 8.

[0089] Furthermore, similar to the example in Figure 6, the reception unit 103A may accept user modifications to the conversion candidates presented by the presentation control unit 107A. This makes it possible to modify the conversion candidates to match the user's intentions before applying them.

[0090] As explained with reference to Figures 8 and 9, the aspect estimation unit 105A may cause the language model M to generate multiple conversion candidates, each abstracting the target area at a different level of abstraction, and the presentation control unit 107A may present the multiple conversion candidates generated by the language model M. The reception unit 103A may then accept a request from the presentation control unit 107A to specify which conversion candidate to use for converting the target area. This provides the effect of generating confidential data in which the target area is abstracted at a level of abstraction desired by the user, in addition to the effects of the information processing device 1.

[0091] Furthermore, as explained with reference to Figure 9, the presentation control unit 107A may present the target location along with the target data, and in response to an operation to specify the presented target location, it may present the concealment method to be applied to that target location, as estimated by the method estimation unit 105A. This provides the effect that, in addition to the effects performed by the information processing device 1, the user can recognize the concealment method to be applied to the target location specified by the user in the target data.

[0092] (Processing flow: Identifying the parts to be concealed) The process flow executed by the information processing device 1A when identifying the area to be concealed will be explained with reference to Figure 10. Figure 10 is a flowchart showing the process flow for identifying the area to be concealed. The flowchart in Figure 10 includes each process of the concealment support method according to this exemplary embodiment.

[0093] In S11 (data acquisition process), the data acquisition unit 101A acquires target data that may contain information to be kept confidential. In S12, the reception unit 103A receives input from the user regarding information to be kept confidential. Note that the process in S12 may be performed before S11, or the processes in S11 and S12 may be performed in parallel. It is also possible to omit the process in S12.

[0094] In S13, the target location estimation unit 102A generates a prompt that instructs the unit to output the information necessary to estimate the location to be concealed in the target data. For example, the target location estimation unit 102A may generate a prompt by inputting the target data obtained in S11 and the information to be concealed received in S12 into a prompt template that instructs the unit to extract the location from the target data that corresponds to the information the user wants to conceal. This allows for the generation of a prompt like prompt 43 in Figure 4, for example.

[0095] In S14 (Target Location Estimation Process), the target location estimation unit 102A estimates the locations in the target data acquired in S11 that should be anonymized, using a machine learning-trained language model M. Specifically, the target location estimation unit 102A inputs the prompt generated in S13 to the language model M. Then, the target location estimation unit 102A estimates the target locations based on the output of the language model M.

[0096] In S15, the display control unit 107A displays the target location estimated in S14 to the user. Then, in S16, the reception unit 103A receives the user's designation of the target location to be concealed from among the target locations displayed in S15.

[0097] In S17 (target location identification process), the target location identification unit 104A identifies the target locations to be concealed in the target data acquired in S11. Specifically, the target location identification unit 104A identifies the target locations specified by the user in S16 from among the target locations presented in S15 as locations to be concealed. This completes the process shown in Figure 10. After the completion of S17, the process shown in S21 in Figure 11 begins.

[0098] Furthermore, the areas estimated in S14 may be designated as areas to be concealed without requiring user confirmation. In this case, S14 becomes an area identification process to identify the areas to be concealed, and the processes from S15 onward are omitted. Then, the process shown in Figure 11, described later, is performed on the identified areas.

[0099] (Processing flow: Generation of confidential data) The processing flow executed by the information processing device 1A when generating confidential data will be explained with reference to Figure 11. Figure 11 is a flowchart showing the processing flow for generating confidential data. The flowchart in Figure 11 also includes the processing of the confidentiality support method according to this exemplary embodiment.

[0100] In S21, the mode estimation unit 105A generates a prompt instructing it to output information necessary to estimate the mode of concealment to be applied to the area to be concealed. This area is the area identified by the area identification unit 104A in S17 of Figure 10. For example, the mode estimation unit 105A may generate a prompt by inputting the target data obtained in S11 of Figure 10 and the area identified in S17 of Figure 10 into a prompt template that instructs the unit to respond with a candidate for conversion of the area to be concealed in the target data. This allows for the generation of a prompt like prompt 71 in Figure 7, for example.

[0101] In S22 (Pattern Estimation Processing), the pattern estimation unit 105A estimates the pattern of concealment to be applied to the target area identified in S17 of Figure 10, using a machine learning-trained language model M. Specifically, the pattern estimation unit 105A inputs the prompt generated in S21 to the language model M. Then, the pattern estimation unit 105A estimates the pattern of concealment based on the output of the language model M.

[0102] In S23, the presentation control unit 107A presents the user with the estimation results from S22, in other words, candidate methods of concealment to be applied to the target area. Then, in S24, the reception unit 103A accepts the user's specification of the method of concealment to be applied to the target area. For example, the reception unit 103A may have the user select one of the candidates presented in S23 to specify the method of concealment to be applied to the target area. As explained with reference to Figure 6, the reception unit 103A may also accept user modifications to the method of concealment presented by the presentation control unit 107A.

[0103] In S25, the concealment unit 106A determines the concealment method to be applied to the target location according to the specifications received in S24. For example, if a conversion candidate is selected for a certain target location in S24, the concealment unit 106A decides to convert that target location to that conversion candidate. If there are multiple target locations, the concealment unit 106A determines the concealment method for each target location.

[0104] In S26, the concealment unit 106A conceals the target portion of the target data in the manner determined in S25 and generates concealed data. Then, in S27, the presentation control unit 107A presents the concealed data generated in S26 to the user. This completes the process shown in Figure 11. Note that presenting the generated concealed data to the user is not mandatory. For example, the concealment unit 106A may store the generated concealed data in the storage unit 11A or the like without having the presentation control unit 107A present it.

[0105] [Reference example] Figure 12 is a block diagram showing the configuration of the information processing device 1B according to this reference example. As shown in the figure, the information processing device 1B includes a target location identification unit 104B and a pattern estimation unit 105B.

[0106] The target location identification unit 104B identifies the target location in the target data that is to be concealed. As with exemplary embodiments 1 and 2, the target data may be any electronic data.

[0107] The method by which the target location identification unit 104B identifies the target location is arbitrary. For example, the target location identification unit 104B may identify a location in the target data specified by the user as the target location. User specifications can be received via an input unit or communication unit (not shown).

[0108] Furthermore, the information processing device 1B may also be provided with the target location estimation unit 102A described in Exemplary Embodiment 2. In that case, the target location identification unit 104B only needs to identify the target location estimated by the target location estimation unit 102A, or the location specified by the user among the target locations estimated by the target location estimation unit 102A, as the target location.

[0109] The mode estimation unit 105B, similar to the mode estimation unit 105A in exemplary embodiment 2, estimates the mode of concealment to be applied to the target location identified by the target location identification unit 104B using a machine learning-trained language model M.

[0110] As described above, the information processing device 1B includes a target location identification unit 104B that identifies target locations in the target data to be concealed, and a mode estimation unit 105B that estimates the mode of concealment to be applied to the target locations identified by the target location identification unit 104B using a machine learning-trained language model M.

[0111] According to the above configuration, since the mode of concealment is estimated using the language model M, it is possible to estimate the mode of concealment according to the content of the target data without having to create in advance something like the concealment rules explained in [Background Technology]. This eliminates or reduces the burden of the task of a person considering and deciding on the mode of concealment, and thus makes it possible to streamline data concealment work. Furthermore, by using the information processing device 1, it becomes possible to optimize the entire data concealment process.

[0112] (Anonymization Support Program) The functions of the information processing device 1B described above can also be implemented by a program. The data concealment support program in this reference example causes the computer to function as a target location identification means for identifying target locations in the target data to be concealed, and as a mode estimation means for estimating the mode of concealment to be applied to the target locations identified by the target location identification means, using a machine learning-trained language model M. This data concealment support program has the effect of making data concealment work more efficient.

[0113] (Methods to support confidentiality) In the data anonymization support method described in this reference example, at least one processor performs a target location identification process to identify the locations in the target data that are to be anonymized, and a mode estimation process to estimate the mode of anonymization to be applied to the locations identified in the target location identification process, using a machine learning-trained language model M. This data anonymization support method has the effect of making data anonymization work more efficient.

[0114] [Variation] The entities executing each process described in the above-described exemplary embodiments and reference examples are arbitrary and not limited to the examples above. For example, a system having the same functions as information processing devices 1, 1A, and 1B can be constructed using multiple devices that can communicate with each other. Also, the entities executing each process shown in the flowcharts of Figures 10 and 11 may be a single device (which can also be called a processor) or multiple devices (which can also be called processors).

[0115] [Examples of implementation using software] Some or all of the functions of the information processing devices 1, 1A, and 1B (hereinafter also referred to as "each of the above devices") may be implemented by hardware such as integrated circuits (IC chips) or by software.

[0116] In the latter case, each of the above devices is implemented, for example, by a computer that executes instructions for a program, which is software that realizes each function. An example of such a computer (hereinafter referred to as Computer C) is shown in Figure 13. Figure 13 is a block diagram showing the hardware configuration of Computer C, which functions as each of the above devices.

[0117] Computer C comprises at least one processor C1 and at least one memory C2. Memory C2 stores a program (confidentiality support program) P that causes computer C to operate as each of the above-mentioned devices. In computer C, processor C1 reads program P from memory C2 and executes it, thereby realizing each of the above-mentioned devices.

[0118] For processor C1, for example, a CPU (Central Processing Unit), GPU (Graphic Processing Unit), DSP (Digital Signal Processor), MPU (Micro Processing Unit), FPU (Floating Point Number Processing Unit), PPU (Physics Processing Unit), TPU (Tensor Processing Unit), quantum processor, microcontroller, or a combination thereof can be used. For memory C2, for example, flash memory, HDD (Hard Disk Drive), SSD (Solid State Drive), or a combination thereof can be used.

[0119] Computer C may also be equipped with RAM (Random Access Memory) for loading program P at runtime and for temporarily storing various data. Furthermore, computer C may be equipped with communication interfaces for sending and receiving data with other devices. Additionally, computer C may be equipped with input / output interfaces for connecting input / output devices such as keyboards, mice, displays, and printers.

[0120] Furthermore, program P can be recorded on a non-temporary, tangible recording medium M that is readable by computer C. Such a recording medium M could be, for example, tape, disk, card, semiconductor memory, or programmable logic circuitry. Computer C can acquire program P via such a recording medium M. Program P can also be transmitted via a transmission medium. Such a transmission medium could be, for example, a communication network or broadcast waves. Computer C can also acquire program P via such a transmission medium.

[0121] Furthermore, each of the above functions of each of the above devices may be implemented by a single processor in a single computer, by multiple processors in a single computer working together, or by multiple processors in each of multiple computers working together. In addition, the programs for implementing each of the above functions in each of the above devices may be stored in a single memory in a single computer, distributed and stored in multiple memories in a single computer, or distributed and stored in multiple memories in each of multiple computers.

[0122] [Additional Notes] This disclosure includes the technologies described in the following appendices. However, the present invention is not limited to the technologies described in the following appendices, and various modifications are possible within the scope of the claims.

[0123] (Note A1) An information processing device comprising: a data acquisition means for acquiring target data that may contain matters to be kept confidential; and a target location estimation means for estimating the parts of the target data that should be kept confidential, using a machine learning-trained language model.

[0124] (Appendix A2) The information processing device according to Appendix A1, comprising a receiving means for receiving input of matters to be concealed, the target location estimation means generates a prompt instructing the extraction of locations corresponding to the matters from the target data, and estimates the target locations based on the output obtained by inputting the generated prompt to the language model.

[0125] (Note A3) The information processing apparatus according to Appendix A1 or A2, comprising: a presentation control means for presenting the target location estimated by the target location estimation means; and a reception means for receiving a designation of a target location to be concealed from among the target locations presented by the presentation control means.

[0126] (Note A4) The information processing apparatus according to Appendix A3, wherein the presentation control means presents conversion candidates for concealing the target portion, and the receiving means accepts the selection or modification of the conversion candidates, and the concealment means either converts the target portion in the target data into the selected conversion candidate to generate concealed data, or converts the target portion in the target data into the modified conversion candidate to generate concealed data.

[0127] (Note A5) An information processing device according to any one of the appendices A1 to A4, comprising a mode estimation means for estimating a mode of concealment to be applied to the aforementioned target portion using the language model.

[0128] (Note A6) The information processing apparatus described in Appendix A5, wherein the aspect estimation means causes the language model to generate a plurality of conversion candidates, each abstracting the target location at a different degree of abstraction, and presents the plurality of conversion candidates generated by the language model; and a receiving means receives a designation of a conversion candidate to be used for the conversion of the target location from among the plurality of conversion candidates presented by the presentation control means.

[0129] (Note A7) The information processing apparatus according to Appendix A5 or A6, comprising a presentation control means that presents the target location together with the target data, and presents a mode of concealment to be applied to the target location, estimated by the mode estimation means, in response to an operation to specify the presented target location.

[0130] (Note A8) An information processing device comprising: a target location identification means for identifying target locations in target data to be anonymized; and a mode estimation means for estimating a mode of anonymization to be applied to the target locations identified by the target location identification means, using a machine learning-trained language model.

[0131] (Note B1) A method for supporting data concealment, comprising: a data acquisition process in which at least one processor acquires target data that may contain information to be concealed; and a target location estimation process in which a machine learning-trained language model is used to estimate the target locations in the target data that should be concealed.

[0132] (Note B2) The anonymization support method according to Appendix B1, wherein at least one processor performs an acceptance process to accept input of matters to be anonymized, and in the target location estimation process, the at least one processor generates a prompt instructing the extraction of locations corresponding to the matters from the target data, and estimates the target locations based on the output obtained by inputting the generated prompt to the language model.

[0133] (Note B3) The concealment support method according to Appendix B1 or B2, wherein at least one processor performs a presentation control process that presents the target locations estimated in the target location estimation process, and a reception process that accepts the designation of target locations to be concealed from among the target locations presented in the presentation control process.

[0134] (Note B4) The concealment support method according to Appendix B3, wherein at least one processor presents conversion candidates for concealing the target portion, accepts selection or modification of the conversion candidates, and performs a concealment process to generate concealed data by converting the target portion in the target data to the selected conversion candidate, or converts the target portion in the target data to the modified conversion candidate to generate concealed data.

[0135] (Note B5) The decryption support method according to any one of the appendices B1 to B4, wherein at least one processor performs an embodiment estimation process that estimates an embodiment of decryption to be applied to the target portion using the language model.

[0136] (Note B6) The privacy protection support method according to Appendix B5, wherein in the manner estimation process, the at least one processor causes the language model to generate a plurality of transformation candidates, each abstracting the target location at a different degree of abstraction; the at least one processor presents the plurality of transformation candidates generated by the language model; and accepts the designation of a transformation candidate from the presented plurality of transformation candidates to be used for transforming the target location.

[0137] (Note B7) The concealment support method according to Appendix B5 or B6, wherein at least one processor presents the target location together with the target data, and in response to an operation to specify the presented target location, it performs a presentation control process that presents a concealment pattern to be applied to the target location, which has been estimated by the pattern estimation process.

[0138] (Note B8) A method for supporting data concealment, comprising: a target location identification process in which at least one processor identifies target locations in target data to be concealed; and a mode estimation process in which a machine learning-based language model is used to estimate the mode of concealment to be applied to the target locations identified in the target location identification process.

[0139] (Note C1) A data concealment support program that enables a computer to function as a data acquisition means for acquiring target data that may contain information that should be concealed, and a target location estimation means for estimating the parts of the target data that should be concealed, using a machine learning-trained language model.

[0140] (Note C2) The anonymization support program described in Appendix C1, wherein the computer functions as a receiving means for receiving input of matters to be anonymized, the target location estimation means generates a prompt instructing the extraction of locations corresponding to the matters from the target data, and estimates the target locations based on the output obtained by inputting the generated prompt into the language model.

[0141] (Note C3) The concealment support program described in Appendix C1 or C2 causes the computer to function as a presentation control means for presenting the target locations estimated by the target location estimation means, and a reception means for receiving the designation of target locations to be concealed from among the target locations presented by the presentation control means.

[0142] (Note C4) The concealment support program as described in Appendix C3, wherein the presentation control means presents conversion candidates for concealing the target portion, the receiving means accepts the selection or modification of the conversion candidates, and causes the computer to function as a concealment means that converts the target portion in the target data into the selected conversion candidate to generate concealed data, or converts the target portion in the target data into the modified conversion candidate to generate concealed data.

[0143] (Note C5) An anti-secrecy support program according to any one of the appendices C1 to C4, which causes the computer to function as an anti-secrecy estimation means for estimating an anti-secrecy method to be applied to the target location using the language model.

[0144] (Appendix C6) The concealment support program described in Appendix C5, wherein the mode estimation means causes the language model to generate a plurality of conversion candidates, each abstracting the target location at a different degree of abstraction, and causes the computer to function as a presentation control means for presenting the plurality of conversion candidates generated by the language model, and a receiving means for receiving a specification of a conversion candidate to be used for converting the target location from among the plurality of conversion candidates presented by the presentation control means.

[0145] (Note C7) The concealment support program described in Appendix C5 or C6, wherein the computer is made to function as a presentation control means that presents the target location together with the target data, and presents a concealment mode to be applied to the target location, which has been estimated by the mode estimation means, in response to an operation to specify the presented target location.

[0146] (Note C8) An anonymization support program that causes a computer to function as a target location identification means for identifying target locations in target data to be anonymized, and a mode estimation means for estimating the mode of anonymization to be applied to the target locations identified by the target location identification means, using a machine learning-based language model.

[0147] (Note D1) An information processing device comprising at least one processor, wherein the at least one processor performs a data acquisition process to acquire target data that may contain information to be made confidential, and a target location estimation process to estimate the target locations in the target data that should be made confidential, using a machine learning-trained language model.

[0148] The information processing device may also include memory. Furthermore, the memory may store a program that causes at least one processor to execute each of the aforementioned processes.

[0149] (Note D2) The information processing apparatus according to Appendix D1, wherein at least one processor performs an acceptance process to accept input of matters to be concealed, and in the target location estimation process, the at least one processor generates a prompt instructing to extract the location corresponding to the matters from the target data, and estimates the target location based on the output obtained by inputting the generated prompt to the language model.

[0150] (Note D3) The information processing apparatus according to Appendix D1 or D2, wherein at least one processor performs a presentation control process that presents the target location estimated in the target location estimation process, and a reception process that accepts the designation of a target location to be concealed from among the target locations presented in the presentation control process.

[0151] (Note D4) The information processing apparatus according to Appendix D3, wherein at least one processor presents conversion candidates for concealing the target portion, accepts selection or modification of the conversion candidates, and performs a concealment process to generate concealed data by converting the target portion in the target data to the selected conversion candidate, or converts the target portion in the target data to the modified conversion candidate to generate concealed data.

[0152] (Note D5) The information processing apparatus according to any one of the appendices D1 to D4, wherein at least one processor performs an embodiment estimation process that estimates an embodiment of concealment to be applied to the target portion using the language model.

[0153] (Note D6) The information processing apparatus according to Appendix D5, wherein in the manner estimation process, the at least one processor causes the language model to generate a plurality of transformation candidates, each abstracting the target location at a different degree of abstraction; the at least one processor presents the plurality of transformation candidates generated by the language model; and accepts the designation of a transformation candidate from among the presented plurality of transformation candidates to be used for the transformation of the target location.

[0154] (Note D7) The information processing apparatus according to Appendix D5 or D6, wherein at least one processor presents the target location together with the target data, and in response to an operation to specify the presented target location, presents the mode of concealment to be applied to the target location, which was estimated in the mode estimation process.

[0155] (Note D8) An information processing device comprising at least one processor, wherein the at least one processor performs: a target location identification process for identifying target locations in target data to be anonymized; and a mode estimation process for estimating modes of anonymization to be applied to the target locations identified in the target location identification process, using a machine learning-trained language model.

[0156] (Note E1) A non-temporary recording medium that records a data secrecy support program that causes a computer to perform a data acquisition process to acquire target data that may contain information that should be kept confidential, and a target location estimation process that estimates the parts of the target data that should be kept confidential using a machine learning-trained language model.

[0157] (Note E2) A non-temporary recording medium that records a concealment support program that causes a computer to perform a target location identification process to identify target locations in target data to be concealed, and a mode estimation process to estimate the mode of concealment to be applied to the target locations identified in the target location identification process, using a machine learning-trained language model. [Explanation of symbols]

[0158] 1. Information Processing Device 101 Data acquisition unit (data acquisition means) 102 Target location estimation unit (target location estimation means) M language model 1A Information Processing Device 101A Data acquisition unit (data acquisition means) 102A Target location estimation unit (target location estimation means) 103A Reception area (reception method) 104A Target location identification unit (means for identifying the target location) 105A Pattern Estimation Unit (Pattern Estimation Means) 106A Secrecy Unit (Secrecy Means) 107A Presentation Control Unit (Presentation Control Means) 1B Information Processing Device 104B Target location identification unit (means for identifying target location) 105B Aspect Estimation Unit (Aspect Estimation Means)

Claims

1. A data acquisition method for acquiring target data that may contain information that should be kept confidential, An information processing device comprising: a target location estimation means for estimating the target locations in the aforementioned target data that should be subject to concealment, using a machine learning-trained language model.

2. It is equipped with a means of receiving input for information that you wish to keep confidential, The information processing apparatus according to claim 1, wherein the target location estimation means generates a prompt instructing the extraction of locations corresponding to the items from the target data, and estimates the target location based on the output obtained by inputting the generated prompt to the language model.

3. A presentation control means that presents the target location estimated by the target location estimation means, The information processing apparatus according to claim 1 or 2, further comprising: a receiving means for receiving the designation of a target location to be concealed from among the target locations presented by the presentation control means.

4. The presentation control means presents conversion candidates for concealing the target area, The receiving means accepts the selection or modification of the conversion candidate. The information processing apparatus according to claim 3, comprising an anonymization means for generating anonymized data by converting the target portion in the target data into the selected conversion candidate, or for generating anonymized data by converting the target portion in the target data into the modified conversion candidate.

5. The information processing apparatus according to claim 1 or 2, further comprising a mode estimation means for estimating a mode of concealment to be applied to the target portion using the language model.

6. The mode estimation means causes the language model to generate a plurality of transformation candidates, each abstracting the target portion at a different degree of abstraction. Presentation control means for presenting the plurality of conversion candidates generated by the language model, The information processing apparatus according to claim 5, further comprising a receiving means for receiving a designation of a conversion candidate to be used for converting the target location from among the plurality of conversion candidates presented by the presentation control means.

7. The information processing apparatus according to claim 5, further comprising a presentation control means that presents the target location together with the target data, and presents a concealment pattern to be applied to the target location, estimated by the pattern estimation means, in response to an operation to specify the presented target location.

8. A means for identifying the target area in the target data that is to be anonymized, An information processing apparatus comprising: an embodiment estimation means for estimating an embodiment to be applied to the target location identified by the target location identification means, using a machine learning-trained language model.

9. At least one processor, A data acquisition process that obtains target data that may contain information that should be kept confidential, A method for supporting data concealment, which includes a target location estimation process that uses a machine learning-based language model to estimate the target locations in the aforementioned target data that should be concealed.

10. Computers, A data acquisition method for acquiring target data that may contain information that should be kept confidential, and A data concealment support program that functions as a target location estimation means, which uses a machine learning-trained language model to estimate the target locations in the aforementioned target data that should be concealed.

Images